Mapping DDI2 to DDI4

Abstract

This poster describes the effort to add a DDI-Codebook (DDI-C) import function into the DDI4R R package. The DDI4 Codebook Group did a lot of the modeling of one section of DDI4 using a spreadsheet mapping DDI-C elements into DDI4 properties. This started with a list of elements used by CESSDA and was refined at the May 2016 Knutholmen Sprint. Unfortunately, these mappings were not always at the leaf node level. An R program also imported DDI-C XML from the European Social Survey and generated a list of unique XPaths of leaf elements used in that set of metadata. These elements, along with corresponding DDI4 leaf paths, were used to update the spreadsheet. This spreadsheet has been further refined to create an actionable table mapping DDI-C leaf values to leaf properties in DDI4. Writing code to import the DDI-C required additional information: • mapping from DDI-C sub-paths to DDI4 Identifiable classes (e.g. all the information for one DDI-C “var” maps to one DDI4 IdentifiableVariable), • mapping abstract target classes to specific extensions, • additional semantic property values like “typeOfMethodology”. Importing DDI-C into a lifecycle level version of DDI like DDI4 also involves identifying repeated metadata like reused value domains (e.g. reused Likert style codelists) that are repeated for multiple variables. An R function served to do this sort of matching using the R “all.equal” function excluding differences in agency, id, and version.